import pandas as pd
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans, DBSCAN, AgglomerativeClustering
from scipy.cluster.hierarchy import dendrogram, linkage

df_aggregation = pd.read_csv("aggregation.csv", sep=";", names=["x1","x2","y"])

df_aggregation

df_aggregation_X = df_aggregation.drop("y", axis=1)
df_aggregation_Y = df_aggregation["y"]

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(2, 11):
    # Criando o objeto KMeans
    kmeans = KMeans(n_clusters=k)

    # Aplicando o algoritmo K-Means
    clusters = kmeans.fit_predict(df_aggregation_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_aggregation_X.iloc[:, 0], df_aggregation_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'K-Means com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(2, 11):
    # Criando o objeto AgglomerativeClustering
    clustering = AgglomerativeClustering(n_clusters=k, linkage='average')

    # Aplicando o algoritmo AgglomerativeClustering
    clusters = clustering.fit_predict(df_aggregation_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_aggregation_X.iloc[:, 0], df_aggregation_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'Average Linkage com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Aplicando o algoritmo de ligação única
clustering = AgglomerativeClustering(n_clusters=7, linkage='average')
clusters = clustering.fit_predict(df_aggregation_X)

# Calculando a matriz de ligação com o algoritmo de ligação média
Z = linkage(df_aggregation_X, method='average')

# Plotando o dendrograma
plt.figure(figsize=(10, 5))
dendrogram(Z)
plt.title('Dendrograma dos Clusters (Average Linkage)')
plt.xlabel('Amostras')
plt.ylabel('Distância')
plt.show()

fig, axs = plt.subplots(3, 3, figsize=(15, 15))

epsilon_values = [0.3, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0]

for k in range(2, 11):

    dbscan = DBSCAN(eps=epsilon_values[k-2], min_samples=8)
    clusters = dbscan.fit_predict(df_aggregation_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_aggregation_X.iloc[:, 0], df_aggregation_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'DBSCAN com eps={epsilon_values[k-2]} e min_samples=8')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

df_aggregation_X = df_aggregation_X.values

clustering = AgglomerativeClustering(n_clusters=7, linkage='average')
clusters_1 = clustering.fit_predict(df_aggregation_X)

dbscan = DBSCAN(eps=1.5, min_samples=8)
clusters_2 = dbscan.fit_predict(df_aggregation_X)

plt.figure(figsize=(10, 5))
plt.subplot(121)
plt.scatter(df_aggregation_X[:,0], df_aggregation_X[:,1], c=clusters_1, cmap='viridis')
plt.title('Average Linkage, n_clusters=7')

plt.subplot(122)
plt.scatter(df_aggregation_X[:,0], df_aggregation_X[:,1], c=clusters_2, cmap='viridis')
plt.title('DBSCAN, eps=1.5, min_samples=8')

plt.tight_layout()
plt.show()

df_d31 = pd.read_csv("d31.csv", sep=";", names=["x1","x2","y"])

df_d31

df_d31_X = df_d31.drop("y", axis=1)
df_d31_Y = df_d31["y"]

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(2, 11):
    # Criando o objeto KMeans
    kmeans = KMeans(n_clusters=k)

    # Aplicando o algoritmo K-Means
    clusters = kmeans.fit_predict(df_d31_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_d31_X.iloc[:, 0], df_d31_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'K-Means com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(25, 34):
    # Criando o objeto KMeans
    kmeans = KMeans(n_clusters=k)

    # Aplicando o algoritmo K-Means
    clusters = kmeans.fit_predict(df_d31_X)

    # Plotando os clusters
    ax = axs[(k-25)//3][(k-25)%3]
    ax.scatter(df_d31_X.iloc[:, 0], df_d31_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'K-Means com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(2, 11):
    # Criando o objeto AgglomerativeClustering
    clustering = AgglomerativeClustering(n_clusters=k, linkage='average')

    # Aplicando o algoritmo AgglomerativeClustering
    clusters = clustering.fit_predict(df_d31_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_d31_X.iloc[:, 0], df_d31_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'Average Linkage com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(25, 34):
    # Criando o objeto AgglomerativeClustering
    clustering = AgglomerativeClustering(n_clusters=k, linkage='average')

    # Aplicando o algoritmo AgglomerativeClustering
    clusters = clustering.fit_predict(df_d31_X)

    # Plotando os clusters
    ax = axs[(k-25)//3][(k-34)%3]
    ax.scatter(df_d31_X.iloc[:, 0], df_d31_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'Average Linkage com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Aplicando o algoritmo de ligação única
clustering = AgglomerativeClustering(n_clusters=31, linkage='average')
clusters = clustering.fit_predict(df_d31_X)

# Calculando a matriz de ligação com o algoritmo de ligação média
Z = linkage(df_d31_X, method='average')

# Plotando o dendrograma
plt.figure(figsize=(10, 5))
dendrogram(Z)
plt.title('Dendrograma dos Clusters (Average Linkage)')
plt.xlabel('Amostras')
plt.ylabel('Distância')
plt.show()

fig, axs = plt.subplots(3, 3, figsize=(15, 15))

epsilon_values = list(range(1,10,1))
for k in range(2, 11):

    dbscan = DBSCAN(eps=(epsilon_values[k-2])/8, min_samples=5)
    clusters = dbscan.fit_predict(df_d31_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_d31_X.iloc[:, 0], df_d31_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'DBSCAN com eps={epsilon_values[k-2]/8} e min_samples=5')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

df_d31_X = df_d31_X.values
clustering = KMeans(n_clusters=31)
clusters_1 = clustering.fit_predict(df_d31_X)

clustering = AgglomerativeClustering(n_clusters=31, linkage='average')
clusters_2 = clustering.fit_predict(df_d31_X)

plt.figure(figsize=(10, 5))
plt.subplot(121)
plt.scatter(df_d31_X[:,0], df_d31_X[:,1], c=clusters_1, cmap='viridis')
plt.title('K-means, n_clusters=31')

plt.subplot(122)
plt.scatter(df_d31_X[:,0], df_d31_X[:,1], c=clusters_2, cmap='viridis')
plt.title('Average Linkager, n_clusters=31')

plt.tight_layout()
plt.show()

df_pathbased = pd.read_csv("pathbased.csv", sep=";", names=["x1","x2","y"])

df_pathbased

df_pathbased_X = df_pathbased.drop("y", axis=1)
df_pathbased_Y = df_pathbased["y"]

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(2, 11):
    # Criando o objeto KMeans
    kmeans = KMeans(n_clusters=k)

    # Aplicando o algoritmo K-Means
    clusters = kmeans.fit_predict(df_pathbased_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_pathbased_X.iloc[:, 0], df_pathbased_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'K-Means com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(2, 11):
    # Criando o objeto AgglomerativeClustering
    clustering = AgglomerativeClustering(n_clusters=k, linkage='average')

    # Aplicando o algoritmo AgglomerativeClustering
    clusters = clustering.fit_predict(df_pathbased_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_pathbased_X.iloc[:, 0], df_pathbased_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'Average Linkage com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Aplicando o algoritmo de ligação única
clustering = AgglomerativeClustering(n_clusters=7, linkage='average')
clusters = clustering.fit_predict(df_pathbased_X)

# Calculando a matriz de ligação com o algoritmo de ligação média
Z = linkage(df_pathbased_X, method='average')

# Plotando o dendrograma
plt.figure(figsize=(10, 5))
dendrogram(Z)
plt.title('Dendrograma dos Clusters (Average Linkage)')
plt.xlabel('Amostras')
plt.ylabel('Distância')
plt.show()

fig, axs = plt.subplots(3, 3, figsize=(15, 15))


epsilon_values = [1, 1.25, 1.5, 1.75, 2, 2.25, 2.5, 2.75, 3]


for k in range(2, 11):

    dbscan = DBSCAN(eps=epsilon_values[k-2], min_samples=10)
    clusters = dbscan.fit_predict(df_pathbased_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_pathbased_X.iloc[:, 0], df_pathbased_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'DBSCAN com eps={epsilon_values[k-2]} e min_samples=10')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

df_pathbased_X = df_pathbased_X.values
clustering = DBSCAN(eps=2.25, min_samples=10)
clusters_1 = clustering.fit_predict(df_pathbased_X)

plt.figure(figsize=(5, 5))
plt.scatter(df_pathbased_X[:,0], df_pathbased_X[:,1], c=clusters_1, cmap='viridis')
plt.title('K-DBSCAN, eps=2.25, min_samples=10')


plt.tight_layout()
plt.show()

df_flame = pd.read_csv("flame.csv", sep=";", names=["x1","x2","y"])

df_flame

df_flame_X = df_flame.drop("y", axis=1)
df_flame_Y = df_flame["y"]

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(2, 11):
    # Criando o objeto KMeans
    kmeans = KMeans(n_clusters=k)

    # Aplicando o algoritmo K-Means
    clusters = kmeans.fit_predict(df_flame_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_flame_X.iloc[:, 0], df_flame_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'K-Means com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Variando o número de clusters de 2 a 10
fig, axs = plt.subplots(3, 3, figsize=(15, 15))

for k in range(2, 11):
    # Criando o objeto AgglomerativeClustering
    clustering = AgglomerativeClustering(n_clusters=k, linkage='average')

    # Aplicando o algoritmo AgglomerativeClustering
    clusters = clustering.fit_predict(df_flame_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_flame_X.iloc[:, 0], df_flame_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'Average Linkage com k={k}')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

# Aplicando o algoritmo de ligação única
clustering = AgglomerativeClustering(n_clusters=4, linkage='average')
clusters = clustering.fit_predict(df_flame_X)

# Calculando a matriz de ligação com o algoritmo de ligação média
Z = linkage(df_flame_X, method='average')

# Plotando o dendrograma
plt.figure(figsize=(10, 5))
dendrogram(Z)
plt.title('Dendrograma dos Clusters (Average Linkage)')
plt.xlabel('Amostras')
plt.ylabel('Distância')
plt.show()

fig, axs = plt.subplots(3, 3, figsize=(15, 15))

epsilon_values = [0.3, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0]

for k in range(2, 11):

    dbscan = DBSCAN(eps=epsilon_values[k-2], min_samples=8)
    clusters = dbscan.fit_predict(df_flame_X)

    # Plotando os clusters
    ax = axs[(k-2)//3][(k-2)%3]
    ax.scatter(df_flame_X.iloc[:, 0], df_flame_X.iloc[:, 1], c=clusters, cmap='viridis')
    ax.set_title(f'DBSCAN com eps={epsilon_values[k-2]} e min_samples=8')
    ax.set_xlabel('Feature 1')
    ax.set_ylabel('Feature 2')

plt.tight_layout()
plt.show()

df_flame_X = df_flame_X.values

clustering = AgglomerativeClustering(n_clusters=4, linkage='average')
clusters_1 = clustering.fit_predict(df_flame_X)

dbscan = KMeans(n_clusters=4)
clusters_2 = dbscan.fit_predict(df_flame_X)

plt.figure(figsize=(10, 5))
plt.subplot(121)
plt.scatter(df_flame_X[:,0], df_flame_X[:,1], c=clusters_1, cmap='viridis')
plt.title('Average Linkage, n_clusters=4')

plt.subplot(122)
plt.scatter(df_flame_X[:,0], df_flame_X[:,1], c=clusters_2, cmap='viridis')
plt.title('KMeans, n_clusters=4')

plt.tight_layout()
plt.show()

	x1	x2	y
0	15.55	28.65	2
1	14.90	27.55	2
2	14.45	28.35	2
3	14.15	28.80	2
4	13.75	28.05	2
...	...	...	...
783	7.80	3.35	5
784	8.05	2.75	5
785	8.50	3.25	5
786	8.10	3.55	5
787	8.15	4.00	5

	x1	x2	y
0	25.0514	5.7475	1
1	26.6614	7.3414	1
2	25.2653	6.2466	1
3	25.2285	4.7447	1
4	25.7529	5.1564	1
...	...	...	...
3095	5.1380	16.6150	31
3096	5.7360	16.9991	31
3097	5.2760	17.2475	31
3098	6.4444	16.2033	31
3099	7.1620	18.1836	31

	x1	x2	y
0	11.25	5.05	1
1	10.95	4.70	1
2	9.85	5.80	1
3	9.80	5.75	1
4	9.15	6.80	1
...	...	...	...
295	17.50	18.05	2
296	16.65	17.50	2
297	15.45	17.30	2
298	15.45	17.80	2
299	15.70	18.40	2

	x1	x2	y
0	1.85	27.80	1
1	1.35	26.65	1
2	1.40	23.25	2
3	0.85	23.05	2
4	0.50	22.35	2
...	...	...	...
235	7.50	26.20	1
236	7.50	25.65	1
237	7.05	25.85	1
238	6.90	27.15	1
239	6.15	26.90	1

Exercícios Aprendizado Nao Supervisionado¶

Autor: Glauco Lauria Marques Filho¶

Resumo¶

Este arquivo contém a resolução dos exercícios da Aula 7 do curso CEDS-808: Aprendizado de Máquina.¶

Sumário¶

Importação de Requisitos¶

1.a. Dataset Aggregation¶

1.a.I Dataset Aggregation - K-means¶

1.a.II Dataset Aggregation - Average Linkage¶

1.a.III Dataset Aggregation - DBSCAN¶

1.a.IV Dataset Aggregation - Melhores Resultados¶

2.a. Dataset D31¶

2.a.I Dataset D31 - K-means¶

2.a.II Dataset D31 - Average Linkage¶

2.a.III Dataset D31 - DBSCAN¶

2.a.IV Dataset D31 - Melhores Resultados¶

3.a. Dataset Pathbased¶

3.a.I Dataset Pathbased - K-means¶

3.a.II Dataset Pathbased - Average Linkage¶

3.a.III Dataset Pathbased - DBSCAN¶

3.a.IV Dataset Pathbased - Melhores Resultados¶

4.a. Dataset Flame¶

4.a.I Dataset Flame - K-means¶

4.a.II Dataset Flame - Average Linkage¶

4.a.III Dataset Flame - DBSCAN¶

4.a.IV Dataset Flame - Melhores Resultados¶